* read_mem and write_mem work now, too
* changed a few things, e.g., handing out DMA address now, instead of id
* works with platform-stress-alloc test (mode 1)
* default implementation should now be working

* changed allocation flags from GFP_DMA (obsolete, discouraged) to
  GFP_KERNEL | __GFP_MEMALLOC
* with >256MB CMA areas this should mean that we do not run into
  ENOMEMs soon, which could be confirmed
* however, DMA allocations tend to destabilize the system, sooner
  or later a memory oops will occur (see platform_stress_alloc)
* not sure how/if to fix that; thought leaving an FPGA-exclusive
  space on top of the DDR via mem= bootarg, but kernel does not
  boot below 500M
* will keep this for now

* minor memory leaks fixed
* destroy sequence fixed
* using DEV logging macros

* continued refactoring: some file renames in device/
* changed deactivate compile flag for performance counters to NPERFC
* re-inserted per-device perfc file with default output
* changed performance counter names and implemented counting
* implemented add function for counters to increase by more than 1
* added filename pattern to tlkm_device_ioctl_cmds.h

* fixed some bugs, mmapping works now
* also improved portability of read/write ioctls

* devices can set their mmap field to provide mmap-support
* reduced log chatter
* removed redundant struct tlkm_dev_cmd
* fixed bug in general read/write via ioctls

* libplatform and libtapasco can be built in the same location now

* default implementations use ioctls, specialized implementations
  may choose a different route per platform
* using deinit everywhere in userland code instead of exit
* several other changes in build configs etc.

* does not fix the issue, though: need a platform-specific function
  to get the status core base address

* new variables for all sorts of dirs, libs, build flags, etc.

* arch/include adn platform/include contain user facing headers
* headers for arch/platform implementation are in common/include

* read and write implemented, as well as bulk commands
* using new device-specific logging macros

* now very common pattern, devices should be distinguishable in syslog
* unifies format for device-specific log entries

* much of the basic initialization can now be generalized and
  re-used for all platforms
* implemented libplatform support for querying, creating and
  destroying devices
* added platform_ctx_t and platform_devctx_t distinction, to
  reflect the platform / device level
* platform_ctx_t holds the refs to the static TLKM interfaces
* platform_devctx_t contains access to dynamic files etc. available
  only after the device has been created
* ported generic modules to new TLKM interfaces
* will need some mechanism to patch-in the board-specific code
* also, will need to implement more calls, e.g., read and write:
  if mmap is available, that is always the best solution; only use
  the TLKM interface, if mmap is not possible
* most of the that code should be generic, too

* adds io-mapping code for GPx and status core
* appears to work, tests require ported libplatform

* interrupt access requires GPx mappings, still TODO
* also need to implement the other ioctl

* problem: udev takes ~10ms to update permissions, can be fixed by waiting
* so far only retrieving info is implemented, rest should be done in platform-specific code